https://www.youtube.com/watch?v=o0Bl3zeDfJM
博客原文地址 https://lilianweng.github.io/posts/2026-06-24-scaling-laws/
先通俗易懂的解读这篇硬核博客 👇
模型越大、数据越多、算得越久，AI 就越聪明——而且变好的速度和规模之间，大致遵循一条"幂律曲线"。
但到底"模型"和"数据"谁该先加大？这是整篇文章争论的核心。
Other newsrooms on this story
· 6 sources
cryptobriefing.com·
Agents-A1 achieves 1T-model performance through long-task training, not bigger parameters
forbes.com·
Stop Measuring AI By Parameter Count. Here’s What Actually Matters
the-decoder.com·
Sina's open model VibeThinker-3B aims to show reasoning compresses well but factual knowledge doesn't
techfundingnews.com·
The data bottleneck every AI startup hits before it scales — TFN
Meituan's LongCat-2.0 shows China can train massive AI models without Nvidia
engineering.fb.com·
Meta’s AI Storage Blueprint at Scale
